Closing the Gap between Cloud Providers and Scientific Users - Cloud Computing with e-Science Applications - page 117

This chapter describes the proposal developed under the e-Clouds project,
which is designed to be a software-as-a-service (SaaS) marketplace for sci-
entific applications running on top of a public cloud infrastructure. It will
include a description of the most important aspects of e-Clouds architec-
ture, emphasizing the different patterns applied for using cloud resources
while hiding the complexity for the end user. A detailed presentation of the
problems faced during development and testing of the first version is also
included. A research group from the Alexander von Humboldt Institute for
Biological Resources (Bogota, Columbia) is used as a case study for testing
some of the ideas outlined and defining the future work for the project.
6.2 What Do Scientific Apps Require?
The concept of cloud computing was mainly developed by having in mind the
usual enterprise web applications. Scientific applications have a particular set
of requirements and characteristics. These new requirements demand a fun-
damentally different approach to problem solving. The next sections discuss
the most important things to consider when relating science and computing.
6.2.1 Flexibility
There is a huge market of general-purpose scientific apps that cover the
day-to-day tasks for the different disciplines. Despite that important offer,
research work often forces technology to adapt, not the other way around.
This is the reason why scientific applications need to be flexible regarding
the kind of processing they support, the input data they receive, and the
outputs they produce. Based on this, multiple file formats must be supported
and a large number of configuration parameters become optional.
Flexibility poses a challenge when porting applications to a cloud platform,
especially when offering them in an SaaS model. An attractive cloud pro-
posal must then include at least the most common configuration options and
a minimum degree of personalization. The way that this can be achieved can
vary greatly between applications; this not only increases the overall com-
plexity but also imposes some restrictions on the solution model.
6.2.2 Platform Maintenance
It is important to consider the wide offer of scientific apps and the multi-
ple platforms in which they can run. This means that part of the migra-
tion to a cloud solution requires deciding under which configuration an
application will run, including operating systems, compilers, and a set of
external libraries, among others. This process becomes more complex when